Making developer experience better. App coming soon.
Joined September 2026
- Tweets118
- Following21
- Followers15
- Likes5
🛠️ Try this: Low when you're deep in tricky logic and want some quiet. High for tests, config and boilerplate where you just want to tab tab tab ⌨️
#VSCode #GitHubCopilot
Follow along. Our app is coming soon 🚀
Copilot finishing your line before you even had the thought? 😅
@code now lets you pick how eager inline suggestions are: Low, Medium or High. They trained one model to handle all three, balancing suggestion size with quality.
xitter.cf/code/status/2114755823…
💡 How proactively should Copilot suggest code? You decide!
With configurable eagerness in VS Code, choose Low, Medium, or High to tailor inline suggestions to your workflow.
Learn how we trained one model to support all three modes and balance suggestion size with quality.
🔗 Read the article: code.visualstudio.com/blogs/…
Happy coding! 💙
Still handing your coding agent a token for your whole account? 👀🔐
@supabase now lets you scope Personal Access Tokens to just the orgs, projects and permissions a job needs. CI or an agent gets one project, not everything.
xitter.cf/supabase/status/211481…
You can now scope your Personal Access Tokens to the organizations, projects, and permissions the job needs. Give a CI pipeline or coding agent access to one project instead of your whole account.
Learn more
supabase.com/blog/scope-your…
🛠️ Try it: if your agent does boring UI work (fill forms, click through a test flow, poke a settings page), switch that step to gpt-6.1-terra and keep your big model for planning.
Same job, smaller bill. Copilot users can pick it in VS Code too.
The numbers they shared 👇
💰 50% lower standard API price than GPT-5.6 Terra
🧪 Matches GPT-6 Astra on Agents' Last Exam (agents using real work software) at 1/10 the price per token
⚡ Over 3x faster on average than Astra and GPT-6.1 Sol
Cheap model that can actually use your computer? Yes please 🖱️💸
@OpenAI shipped GPT-6.1 Terra. It reads the screen, clicks across apps and takes action. It's live in Codex and ChatGPT Work, and in the API as gpt-6.1-terra.
xitter.cf/OpenAI/status/21147807…
Our take: an agent that asks one good question beats one that guesses fast 🤝
Which tool would you make async first? 👀
#MCP #ClaudeCode
Follow along. Our app is coming soon 🚀
🛠️ Try it if you build MCP servers: pick your slowest tool (a big export, a crawl, a full test run) and make it a Task, so it returns right away and reports back later.
Then add one Elicitation before anything risky, like a delete or a payment.
Elicitations are the other half 👇
A plugin can stop mid-task and ask you for input, like confirming a refund or filling in a missing detail.
No more guessing a value and failing at step 7 of 9 😅
Live today in the Claude Code CLI. More Claude products soon.
Your MCP plugin doesn't have to finish in one breath anymore ⏳🔌
@ClaudeDevs added MCP Tasks and Elicitations to Claude. Tasks let Claude hand a plugin a long job, like deep research, and get the result when it's done.
xitter.cf/ClaudeDevs/status/2113…
Fixed versions 👇
🟢 Node.js 22.23.4, 24.22.1, 26.12.1
🟢 Bun 1.4.4
🟢 Envoy 1.40.0, or the latest patch on your branch
Then rebuild your Docker images and restart every worker and cron job. An old container keeps the old runtime 😅
Run Node.js, Bun or Envoy? Go patch today 🩹
@OpenAI found a remote code execution bug in Brotli while red-teaming their training setup with agents. A bad compressed payload could take over a server, or a client like curl --compressed 😬
xitter.cf/OpenAI/status/21137663…
During this red-teaming exercise, we also found a vulnerability that allows for unauthenticated remote code execution in Brotli, a widely-deployed compression library included in many pieces of software such as Node.js, Bun, and Envoy.
The flaw could let an attacker run their own code on affected systems by sending or replying with malicious compressed data, without any need to create a user account. That is, an unpatched server that accepts compressed input could be taken over by a malicious client, and an unpatched client that accepts compressed input (such as “curl --compressed”) could be taken over by a malicious server.
This root cause issue was actually fixed in the Brotli source code in February, but it wasn’t known at the time that it gave rise to a security vulnerability and there was no more recent release of Brotli. Since Brotli is built into other software, any software which includes this library needs to take the update too, followed by the users of that software. We disclosed the vulnerability to Brotli and other downstream projects such as Node.js, which released or are currently releasing security updates, and Brotli cut new release versions and published a CVE.
More details on the patch and mitigations below:
patchtheplanet.org/brotli
How it works 👇
🔎 Regex casts a wide net for anything key-shaped
🧠 The model reads the file path, var names and nearby code
🚫 "example", "dummy" and test fixtures get filtered out
Fast enough that generic secrets now get caught at push time, not after.
A tiny model just beat a big LLM at catching leaked keys 🔑🤏
@github swapped its LLM secret scanner for a small fine-tuned ModernBERT classifier. 11.5% more accepted alerts, 86% faster scans, 98% cheaper to run.
xitter.cf/github/status/21144796…
Faster secret scanning. Less noise. GitHub's classifier raised acceptance by 11.5% and cut median scan time by 86%.
github.blog/ai-and-ml/github…
haha yeh 100% 😄 happy path is the easy demo. permission popups and slow animations are where real testing lives. would love agents that wait for the screen to settle before they tap